Apple is an existential threat to the PC

[iroboto]

A long time ago Steve Jobs said Post-PC era world.
Yup, it's a long time coming, but the writing was on the wall and he wasn't afraid to sell it. There will always be a need for PCs, just not like the hey day it used to be.

That's being said, in reference to OP, I'm not sure it's necessarily Apple that's the issue. The world has just moved onto smaller form factors because over time those form factors are more comfortable in the hands on the users that have grown up with touch screen devices. Children who are born 2007+

 

[MfA]

Pick your poison.

I'm greedy, I want my preferred platform to also have devices which Just Werk. I want it all, I want it now. I think without, PC gaming and the PC industry in general are not long for this life.

 

[Entropy]

I’m still convinced that this Bloomberg rumor was planted and funded by someone looking to short Intel stock.
Makes for interesting discussion fodder, but just because Apple is capable doesn’t mean that they will. Having a contingency plan makes sense, just as in their mobile GPUs, but doesn’t mean that it will ever materialize into product as long as intel provide sufficient value. (That is, letting intel feel that they are replaceable as a supplier is as important as having the capability to replace them.)

 

[Laurent06]

There's one more thing to think about in this discussion, IMO.

If Apple moves away from x86, Apple stands to lose market share as many professionals only consider a Mac because they can also easily run x86 Windows applications if needed for work.

That's not what I see at work. Many people don't need Windows any more on their Mac. That was not the case some years ago.

 

[Laurent06]

Last year Microsoft said Windows on ARM would run 80x86 code at "near native speeds" and explained how. Emulation isn't the performance hit it once was, well it's still sub-optimal which is why you generally don't do traditional emulation, you cross-compile or build a new binary once, then run that native code. Simples.

Here's a geekbench comparison of HP ENVY X2 vs a phone based on Qualcomm SD835: http://browser.geekbench.com/v4/cpu/compare/7831958?baseline=7798443

The price to pay for emulation is quite high: it's more than twice slower.

 

[Entropy]

Here's a geekbench comparison of HP ENVY X2 vs a phone based on Qualcomm SD835: http://browser.geekbench.com/v4/cpu/compare/7831958?baseline=7798443

The price to pay for emulation is quite high: it's more than twice slower.

That wouldn’t be the case for much code if Apple switched their Macs to ARM though. All the system software would be native, all Apple apps, all "live" software that is actually supported, anything that is submitted using bitcode (hmm, wonder why they’ve been promoting that for apps!) - how much is left, really? That is performance critical?
Seriously, in the overall picture, I just can’t see this as being an issue.
Since you are a knowledgeable chap, just how would you evaluate Apples current A11 vs. Intels core stuff at ISO power? (In 2020, on TSMC N5 Apples process is quite possibly in a better place vs. Intel than at present, and presumably they aren’t staying static on their core designs either, so the current ratio could serve as some kind of lower bound.)

 

[zed]

That's not what I see at work. Many people don't need Windows any more on their Mac. That was not the case some years ago.

yes anecdotally for myself, theres stuff I do now on the web that I used to do with apps, in fact I'll hazzard a guess even though I use a PC ~80 hours a week, I've only used maybe 10 different programs this year

 

[iMacmatician]

In terms of Apple going ARM for their MacOS systems, there are tangible advantages for Apple in terms of tailoring their SoC to the hardware they want to build whether in features, performance, power draw, timing of introduction, or cost. But this has to translate into an equally tangible benefit for their customers, or the appeal of their machines won’t increase.

I read this tweet last year which puts an interesting perspective on a transition from a perspective external to Apple.

Apple will start the switch to ARM not from Macbooks or the Mac Mini, but from the Mac Pro I expect.

Apple needs to break the "ARM is slower" myth at all costs.

I think it's a good idea for Apple to have a pro desktop (iMac Pro or Mac Pro) among the first Macs to transition to ARM. In my opinion, this is due to a number of problems that Apple faces:

1. Apple needs to show that it can release a suitable ARM-based CPU for a high end Mac.
   • The general assumption regarding a hypothetical Apple-ARM switch is that Apple will start with a product like the MacBook and later transition the higher power Macs.
   • A "bottom-up" transition may give the impression that ARM = lower end and Intel = higher end.
     • Note that this is already the case when Apple's entire computer lineup from Apple Watch to Mac desktop is considered.
     • The iPad Pro is a mobile device while the Surface Pro and MacBook are considered to be "real" computers, and this divide would still hold even if the iPad Pro CPU was the fastest of the three. My concern is that if Apple starts the transition with (edit: only) a low-TDP Mac with an A[#]X SoC (for example) and if there are any nontrivial software difficulties, then the ARM Mac may be perceived as being on the wrong side of the divide.
     • A high end ARM Mac that is more powerful than any Intel Mac, even in a limited set of "pro" software, wouldn't give the same impression.
2. Apple needs to show that its high end ARM-based CPU compares favorably with high end Intel and AMD CPUs.
   • The cylinder Mac Pro was not updated for more than three years after its debut, and eventually Apple held a meeting last year with some reporters to reassure people of its commitment to the pro market.
   • So it's not a good idea for Apple to perform a relative downgrade (e.g. skipping Sandy Bridge to go from 2 Westmere CPUs in the tower Mac Pro to 1 Ivy Bridge CPU in the cylinder Mac Pro, or the PowerBook G4 speculation at the end of this page) just to make the ARM Mac look better.
3. Benchmark comparisons across architectures are met with skepticism.
   • Geekbench is very popular but is considered to be a poor benchmark for cross-architectural comparisons (whether that criticism is accurate or not).
   • Any benchmark that shows "ARM CPU > Intel CPU" will be (and often is already) doubted.

The following hypothetical comparison is probably a good way to convince both consumers and professionals that an ARM switch is the way to go, at least on the hardware side.

• Specs and benchmarks of an ARM-based Apple CPU compared to the highest-end Intel and AMD workstation CPUs:
  • More cores (because people like more cores).
  • Higher single-threaded performance (an average that is anything outside the margin of error) in SPEC CPU 2017.
  • Significantly higher multithreaded performance ( ≥ 50% average) in SPEC CPU 2017.
  • If the Apple CPU is the "highest performing single- and multi-threaded compute core" in a "serious" benchmark such as SPEC, then problem #2 should be mitigated.
• A pro ARM Mac compared to pro Intel Macs in real-world scenarios:
  • Higher overall performance in a wide range of consumer software.
  • Higher overall performance in a wide range of professional software (e.g. Final Cut Pro).
  • Demos of the above, not just numbers on slides.
  • The above should address problem #1.
  • If/when emulation is used then the ARM Mac should not be too far behind its Intel-based predecessor (according to Apple, there will be a new Mac Pro in 2019, so that could be a comparison point).

Anything significantly less and a not-insignificant proportion of users, especially "pro" users, may continue to doubt whether Apple's own CPUs can successfully replace Intel in the entire Mac line.

 

[Entropy]

I read this tweet last year which puts an interesting perspective on a transition from a perspective external to Apple.

I think it's a good idea for Apple to have a pro desktop (iMac Pro or Mac Pro) among the first Macs to transition to ARM. In my opinion, this is due to a number of problems that Apple faces:

1. Apple needs to show that it can release a suitable ARM-based CPU for a high end Mac.
   • The general assumption regarding a hypothetical Apple-ARM switch is that Apple will start with a product like the MacBook and later transition the higher power Macs.
   • A "bottom-up" transition may give the impression that ARM = lower end and Intel = higher end.
     • Note that this is already the case when Apple's entire computer lineup from Apple Watch to Mac desktop is considered.
     • The iPad Pro is a mobile device while the Surface Pro and MacBook are considered to be "real" computers, and this divide would still hold even if the iPad Pro CPU was the fastest of the three. My concern is that if Apple starts the transition with a low-TDP Mac with an A[#]X SoC (for example) and if there are any nontrivial software difficulties, then the ARM Mac may be perceived as being on the wrong side of the divide.
     • A high end ARM Mac that is more powerful than any Intel Mac, even in a limited set of "pro" software, wouldn't give the same impression.
2. Apple needs to show that its high end ARM-based CPU compares favorably with high end Intel and AMD CPUs.
   • The cylinder Mac Pro was not updated for more than three years after its debut, and eventually Apple held a meeting last year with some reporters to reassure people of its commitment to the pro market.
   • So it's not a good idea for Apple to perform a relative downgrade (e.g. skipping Sandy Bridge to go from 2 Westmere CPUs in the tower Mac Pro to 1 Ivy Bridge CPU in the cylinder Mac Pro, or the PowerBook G4 speculation at the end of this page) just to make the ARM Mac look better.
3. Benchmark comparisons across architectures are met with skepticism.
   • Geekbench is very popular but is considered to be a poor benchmark for cross-architectural comparisons (whether that criticism is accurate or not).
   • Any benchmark that shows "ARM CPU > Intel CPU" will be (and often is already) doubted.

The following hypothetical comparison is probably a good way to convince both consumers and professionals that an ARM switch is the way to go, at least on the hardware side.

• Specs and benchmarks of an ARM-based Apple CPU compared to the highest-end Intel and AMD workstation CPUs:
  • More cores (because people like more cores).
  • Higher single-threaded performance (an average that is anything outside the margin of error) in SPEC CPU 2017.
  • Significantly higher multithreaded performance ( ≥ 50% average) in SPEC CPU 2017.
  • If the Apple CPU is the "highest performing single- and multi-threaded compute core" in a "serious" benchmark such as SPEC, then problem #2 should be mitigated.
• A pro ARM Mac compared to pro Intel Macs in real-world scenarios:
  • Higher overall performance in a wide range of consumer software.
  • Higher overall performance in a wide range of professional software (e.g. Final Cut Pro).
  • Demos of the above, not just numbers on slides.
  • The above should address problem #1.
  • If/when emulation is used then the ARM Mac should not be too far behind its Intel-based predecessor (according to Apple, there will be a new Mac Pro in 2019, so that could be a comparison point).

Anything significantly less and a not-insignificant proportion of users, especially "pro" users, may continue to doubt whether Apple's own CPUs can successfully replace Intel in the entire Mac line.

I think this is a good analysis insofar that if Apple is going to shift to their own CPU/GPU solutions, it would be important that they bring tangible improvement, and I agree that power draw and GPU performance for portable computers alone wouldn’t really cut it. (They are already there basically, so that wouldn’t take too much effort on their part.) That was why I tried to tease a bit of an estimate out of Laurent in terms of performance vs. Core at the same power draw, because my reasonably researched opinion is that overall, the A11 is already a fair bit more efficient than intels offerings. Add that Apple by all appearances will have access to a lithographic process that is relatively better placed vs. intel than the situation is for the A11, and the "better single threaded performance" should be eminently doable. Better multithreaded performance is comparatively easy if you’re willing to throw silicon area at the problem. And Apple can do that, they pay a tidy sum for the intel chips they put into their systems, if viewed from a $/mm2 perspective. Hell, even AMD can do that, right now, with miniscule resources, a sub par process, and tied to the same x86 cruft that intel is.

Apples new systems would need to make a splash. If they do, Apple could also manage to increase their Mac marketshare vs. Wintel, bringing economy of scale benefits. It would be a bold move, one that could rock the boat in personal computing a bit, and Apples role in it a lot.

I’d love to see it for many reasons, and the ground is prepared for it, but I’m not sure it will actually happen. This particular wolf has been cried for a long time, and to the best of my knowledge there is no source whatsoever for this claim, hence my "shorting intel stock" assumption. No leak out of TSMC, or any IC design tool folks who you might otherwise assume could have access to early indications. Nothing. So....?

 

[MfA]

No one working on the OS-X Arm port has talked, we know for a fact it exists ... but nothing else. So I don't see why a processor would be any different.

They are pretty mercenary about information security.

 

[tuna]

No one working on the OS-X Arm port has talked, we know for a fact it exists ... but nothing else. So I don't see why a processor would be any different.

They are pretty mercenary about information security.

Isn't the MacOS kernel the same as the iOS kernel? They shouldn't really have to port anything except some boot loader utilities.

 

[Entropy]

Isn't the MacOS kernel the same as the iOS kernel? They shouldn't really have to port anything except some boot loader utilities.

Yes it is. It’s even available on GitHub in ARM version for the first time. (Article link.)
Craig Federighi described how it works like this a year ago:

For our developers, we think we want a common platform at the lower level for a developer to build an app that could work on iOS, even tvOS, MacOS. The differentiated bits are the higher level user environment and some of the highest level frameworks that target that user experience.

Every shred of evidence points to Apple having full MacOS running on ARM. The question is merely if they will use it for products. The ground work is done, even when it comes to apps, with their emphasis on LLVM/BitCode.
As I said earlier, I can’t see running MacOS and apps on ARM being problematic at all, and there are advantages to adding underlying hardware to their control of the full package (interview mentioning pro workflow team examples of platform control advantages).
The only thing that might be an issue is the performance of Windows emulation. Laurent posted benchmark numbers indicating a factor of two penalty for emulation. That’s actually not too bad assuming good native performance, and more intrigueing is running Windows on ARM natively. Microsoft sells software licenses, and if people choose to run them on Macs, that’s fine by them. Opens some interesting possibilities for the future for sure.

Will Apple go ARM for MacOS products though? Well, we just can’t know. It would be interesting.

 

[Entropy]

While this is delightful speculation fodder, the unknowns are legion. Assuming 2020, that is when TSMC N5 volume production is scheduled to start. So that would give us some ballpark idea about density if used. However, what would those CPU cores look like? The GPU? Other co-processors/functional blocks? The memory subsystem? The advantage for Apple is that this would be entirely up to their discretion, but it leaves an armchair speculator floundering in a sea of possibilities.

 

[Mobius1aic]

"Apple as expected is ditching Intel x86 processors...........but in an industry wide shocker............for MIPS"

Joking aside.........I think the rest of the industry would be leaking any Apple ARM news at this point if the conversion was from the top-down. No way Apple is going to expect everyone who produces either hardware or software for pros and prosumers to not be ready for a transition. The sooner and smoother other entities are ready, the easier it will be for consumers. External devices and discreet GPUs will need drivers, unless Apple's new graphics IP is ready and arrives with an ARM transition to kick AMD out of the game. That makes a poop load of sense actually.................

 

[Laurent06]

Since you are a knowledgeable chap, just how would you evaluate Apples current A11 vs. Intels core stuff at ISO power?

I'm not knowledgeable at all in implementation. And if I had seen non-public comparisons I could not share them

My feeling is that Apple pushed for very high IPC, and usually this is not very good for power and/or efficiency. But at the same time, the best efficiency point is not very interesting if you need performance. Yeah very generic remarks, sorry I don't know more about Apple...

 

[Cyan]

The software platforms are fundamentally different in how they are funded. Android is provided ”for free” - it is basically ad-ware. Windows is funded by licensing. MacOS/iOS is funded by hardware sales, so only allows/supports running the OS on their own devices. Pick your poison.
It influences the feel of the respective platforms though in terms of breadth of hardware choice, longevity of support, coherence et cetera.

In terms of Apple going ARM for their MacOS systems, there are tangible advantages for Apple in terms of tailoring their SoC to the hardware they want to build whether in features, performance, power draw, timing of introduction, or cost. But this has to translate into an equally tangible benefit for their customers, or the appeal of their machines won’t increase. Would the benefits of rolling their own outweigh the advantages of just buying off the shelf PC components and put their engineering into design, integration and software? Weeell - maybe. Impossible to say really without having the actual products and pricepoints. It would be really interesting to see. As a long time mac user, I’ve already gone through three different underlying ISAs, and another change isn’t such a big deal. I’m more curious as to what it would enable, than fearful of dire consequences, because I know there wouldn’t really be any.

I can’t see any red blooded tech enthusiast not be curious about what Apple could do on TSMC 5nm in 2020 if given 150mm2 and 50W.

Overall I think the current technological world is going through an existencial crisis. Windows is no longer the behemoth it used to be, and to be honest, it is for good, nor it is a standard the monolithic approach of selling several versions of Windows. It was too orthodox for most people, just like Ballmer.

ARM 64 processors are going to work on Windows soon. https://www.windowscentral.com/windows-10-arm-add-64-bit-app-support-may

This leaves the ball on Intel's roof. I heard that Intel is copying ARM architecture for computers in the 10th generation. So I wouldn't write them off.

 

[Deleted member 13524]

ARM 64 processors are going to work on Windows soon. https://www.windowscentral.com/windows-10-arm-add-64-bit-app-support-may

The only news from that is devs will be able to compile for 64bit Armv8, not that ARM CPUs will be able to emulate x86-64 applications.

I didn't even know they couldn't compile to ARM's native 64bit instructions already...
Windows on ARM is even messier than I thought...

 

[rcf]

Apple is an existential threat to its original core of professional Macintosh users.

 

[Silent_Buddha]

Apple is an existential threat to its original core of professional Macintosh users.

LoL, what a clusterfuck. To summarize.

• They wanted to pay for repairs due to their own carelessness.
• The Apple store started the process.
• Apple headquarters won't approve sending the parts to the Apple store to repair the iMac Pro.
• They are told they need to go to an Apple Certified repair place.
• The Apple Certified repair place can fix it and Apple Headquarters will send them the parts, if the repair place has someone with Apple Pro certification.
• Apple Pro certification does not currently exist.

What. The. Fuck?

Regards,
SB

 

[BRiT]

Think Differently #Apple